Meta-learning: Towards Fast Adaptation in Multi-Subject EEG Classification

Multi-subject electroencephalography (EEG) classification involves building a model for automatically categorizing brain waves measured from multiple subjects who undergo the same mental task. A huge amount of subject-dependent variability exists in EEG data. Thus, subject-to-subject transfer or a quick adaptation of the model to a small amount of new data is important to achieve a satisfactory performance in multi-subject EEG classification. Recent advance in meta-learning has a great potential in multi-subject EEG classification. In this paper, we introduce recent development of meta-learning, emphasizing its role to enable the model to quickly adapt to a novel task.